State Of Charge Estimation And State Of Health Prediction Of Lithium-ion Power Batteries Under Dynamic Conditions

As the "heart" of electric vehicles,batteries are currently the research focus of new energy vehicles.The state of charge(SOC)of a lithium ion battery can reflect the current remaining capacity of the battery,which can be used to calculate the remaining mileage of the battery car,while the state of health(SOH)of the battery reflects the remaining life of the battery.How to improve the prediction accuracy of battery SOC and SOH as well as achieve real-time online prediction is currently the focus of research on battery management systems for electric vehicles.In this paper,the optimized neural network algorithm can achieve accurate estimation of battery SOC,and the optimized Gaussian process regression algorithm can well achieve accurate prediction of battery SOH.This article first describes the basic structure of commonly used lithium ion power batteries,and analyzes the basic principles of battery operation.At the same time,the meaning and functions of several important parameters of lithium ion batteries are emphatically introduced.Finally,the factors affecting the SOC and SOH of lithium batteries were analyzed.Secondly,this article adopts dynamic condition experiments to replace the constant current and constant voltage charging and discharging experiments commonly used in previous studies.Dynamic condition better simulates the discharge logic of electric vehicles during operation,making the subsequent algorithm predictions in this article more practical.The dynamic stress test cycle and the federal urban driving plan cycle were used respectively.A single dynamic operating cycle experiment was conducted for SOC and three types of data were collected,including current,voltage,and temperature.A 300 cycle aging experiment was conducted for SOH and current and voltage data were collected.In this paper,the BP neural network has been upgraded.Three combined algorithms,namely,traditional BP neural network,genetic algorithm combined with BP neural network,and thought evolution algorithm combined with BP neural network,have been used to predict SOC,and four evaluation indicators have been used to evaluate the accuracy of the algorithm.The results show that the combination of thought evolution algorithm and BP neural network has achieved good results,with the error controlled within 2% under DST conditions and within 1.5% under FUDS conditions.Finally,the data set obtained during the battery cycle aging test was preprocessed,and the constant pressure differential charging time was extracted from the constant current charging stage by correlation analysis method as a health factor for predicting the battery SOH.A Gaussian process regression algorithm was adopted to predict the SOH of the lithium ion battery.A Gaussian process regression prediction model was established using combined kernel functions and maximum likelihood method to optimize the superparameters,and the constant pressure differential charging time was used as an input,The residual capacity of lithium ion batteries is predicted using SOH as the output.The results show that the method can effectively predict the SOH of batteries.